The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted third-party content via image URLs provided by the user. While the skill includes documentation on mitigations and uses JSON structures to separate commands from data, the ingestion of external data remains a known attack surface.
Ingestion points: External URLs passed to the image_urls array within the JSON input to the runcomfy command.
Boundary markers: The skill instructions emphasize passing inputs as a JSON string to the --input flag to prevent shell expansion and interpolation issues.
Capability inventory: The skill utilizes the Bash(runcomfy *) tool to execute outpainting tasks and manage local file downloads of resulting images.
Sanitization: Inputs are structured as JSON; however, the skill does not perform active sanitization of the visual data contained within the external images.
[COMMAND_EXECUTION]: The skill invokes the runcomfy CLI tool through the Bash tool to interact with the RunComfy API. This usage is restricted to the specific CLI interface and is standard for the skill's intended purpose.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the @runcomfy/cli package from the official NPM registry. This is the primary driver for the skill and is a standard development practice.

image-outpainting